3.1 Changes in meteorological variables during the three periods
The daily average temperature, wind speed (WS), wind direction (WD), and relative humidity (RH) from January 1st, 2020 to May 20th, 2020 are presented in Fig. 2. Comparing the PRE period to the LOOK period, lower temperatures were observed, while the highest temperatures were recorded during the POST period. The WS exhibited a similar pattern of change. However, during the LOOK period, there was a higher frequency of north winds (337.5–22.5°) as indicated by the WD. Conversely, the PRE period experienced relatively calm conditions with greater variability in wind direction. The PRE and POST periods displayed the most stable and highest RH values, respectively, compared to the other cycles where RH fluctuated. The mean and standard deviations for temperature, WS, and RH were − 2.25 ± 5.86℃, 8.14 ± 3.63km/h, and 59.81 ± 13.92%, respectively (Table 1). Northeast winds (337.5–90°) and south winds (135–225°) were the most common WD during all three periods. The daily meteorological variables for 2019 and 2021 were also compared to the lockdown period (2020) (Table 1) and illustrated in Fig. S1-S2. The average WD for 2019 and 2021 remained predominantly south without significant changes. There were no significant differences in the other variables, except for slightly lower RH in 2019 compared to the subsequent two years. Therefore, there were no substantial variations in meteorological variables among different years.
Table 1
Basic statistics of meteorological variables for the LOCK period from 2019 to 2021.
| Parameter | Number | Unit | Mean | Std Dev | Min | Max |
2019 | Temperature | 57 | ℃ | -1.84 | 6.84 | -15.8 | 12.3 |
Relative humidity | 57 | % | 44.35 | 13.35 | 20 | 77 |
Wind direction | 57 | ° | 189.95 | 104.17 | 1 | 357 |
Wind speed | 57 | km/h | 7.93 | 3.37 | 3 | 15 |
2020 | Temperature | 58 | ℃ | -2.25 | 5.86 | -14 | 8.5 |
Relative humidity | 58 | % | 59.81 | 13.92 | 30 | 93 |
Wind direction | 58 | ° | 169.52 | 119.82 | 0 | 359 |
Wind speed | 58 | km/h | 8.14 | 3.63 | 2 | 17 |
2021 | Temperature | 57 | ℃ | -2.35 | 6.41 | -17 | 6.2 |
Relative humidity | 57 | % | 55.25 | 14.59 | 25 | 92 |
Wind direction | 57 | ° | 172.91 | 117.95 | 1 | 357 |
Wind speed | 57 | km/h | 9.00 | 3.35 | 3 | 18 |
3.2 Characterization of air quality changes
3.2.1 Temporal characterization
According to the "Technical Regulations for Air Quality Index (AQI) (Trial Implementation)" (HJ633-2012), the air quality in Shenyang during the epidemic was classified. According to Table 2, In January 2020, the average and peak AQI values were 141.22 and 315, respectively. The number of days with air quality at or above the level of severe pollution (AQI > 200) accounted for 22.58%, which was higher than in 2019 and 2021, indicating an overall poor air quality situation. However, after the Level I response was activated, there was a significant improvement in air quality. The average AQI values in February and March 2022 were 75.34 and 65.45, respectively, showing a notable decrease compared to January (p < 0.01). The peak values during these months were 208 and 129, and the number of days with severe pollution or above accounted for 3.45% and 0%, respectively. This indicates a considerable reduction in pollution levels. Figure 3 shows that the AQI values remained relatively stable from February 12th to March 22nd. Additionally, although there was no significant difference in the average AQI values during the lockdown period (80.17) compared to the reference periods of 2019 (97.82) and 2021 (92.46), there was a higher percentage of excellent air quality days at 77.59% during the lockdown period. This suggests that, despite similar average AQI values, the air quality during the lockdown period was generally better. It is worth mentioning that the AQI values were higher on January 25th and January 29th in 2020, possibly due to inadequate control of fireworks during the Spring Festival and poor atmospheric diffusion conditions. After March 24th, with the implementation of Level I epidemic prevention and control measures and the resumption of work and production in Shenyang, the AQI rebounded. The average AQI value in April (97.13) showed a significant increase compared to March (65.45) ((p < 0.05). Therefore, it can be inferred that the improvement in air quality during the epidemic is related to the preventive and control measures adopted in China. During the period from January 26th to March 23rd, residents reduced non-essential outings, and mobile and fixed sources of pollution in Shenyang and its vicinity decreased, leading to a reduction in pollutant emissions and an improvement in air quality. These measures played a crucial role in mitigating pollution levels during the epidemic period.
Table 2
Comparison of air quality in different periods of Shenyang.
Phase | January | February | March | April | May | During LOCK |
2019 | Average AQI | 101.16 | 97.18 | 92.90 | 84.40 | 83.29 | 97.82 |
Peak AQI | 262 | 227 | 251 | 374 | 197 | 251 |
Number of excellent and good days (%) | 61.29 | 71.43 | 77.42 | 83.87 | 77.42 | 73.68 |
Number of severe and above days (%) | 6.45 | 3.57 | 9.68 | 3.23 | 0 | 7.02 |
2020 | Average AQI | 141.22 | 75.34 | 65.45 | 97.13 | 67.48 | 80.17 |
Peak AQI | 315 | 208 | 129 | 173 | 127 | 257 |
Number of excellent and good days (%) | 31.48 | 75.86 | 93.55 | 60.00 | 83.87 | 77.59 |
Number of severe and above days (%) | 22.58 | 3.45 | 0 | 0 | 0 | 6.90 |
2021 | Average AQI | 88.29 | 77.61 | 104.97 | 70.93 | 72.81 | 92.46 |
Peak AQI | 275 | 142 | 247 | 142 | 139 | 275 |
Number of excellent and good days (%) | 80.65 | 78.57 | 51.61 | 90.00 | 80.65 | 64.91 |
Number of severe and above days (%) | 9.68 | 0 | 6.45 | 0 | 0 | 5.26 |
3.2.2. Spatial characterization
Using the inverse distance weighted spatial interpolation method in ArcGIS 10.8, we interpolated the monthly average Air Quality Index (AQI) values of 11 air quality monitoring stations in Shenyang City from January to April between 2019 and 2021. This allowed us to create a comparative map showing the spatial distribution of AQI values in the urban area of Shenyang City for the same months across different years. Detailed information on the locations of the monitoring stations and the monthly average AQI values from January to April for the years 2019 to 2021 can be found in Tables S1-S3. Figure 4 illustrates the range of monthly average AQI values in Shenyang City from January to April 2020. The values ranged from 115 to 164 (Fig. 4a) in January, 63 to 87 (Fig. 4b) in February, 60 to 74 (Fig. 4c) in March, and 81 to 118 (Fig. 4d) in April. Overall, the distribution trend of air quality in Shenyang City remained consistent during the study period. The southwestern part of the city, including Yuhong District, Tiexi District, Sujiatun District, and the western part of Hunnan District, exhibited relatively higher overall AQI values. On the other hand, the northeastern part of the city, comprising Shenbei New District and the northern part of Hunnan District, had relatively lower overall AQI values. Monitoring stations situated in industrial and densely populated areas, such as Tiexi District and Hunnan District, recorded higher AQI values compared to the ecological tourist area in the northeastern part of Shenyang, namely Shenbei New District. This pattern aligns with the general understanding that AQI values tend to be higher in industrial areas than in ecological tourist areas, demonstrating how different industrial structures contribute to regional variations in AQI. Notably, the implementation of strict control measures during the COVID-19 pandemic led to the closure of most industrial enterprises in Shenyang, resulting in a reduced AQI gap between the north and south regions. Analyzing different time periods, we observed no significant fluctuations in the monthly average AQI values for January (87–110), February (79–112), and March (78–107) during the early stages of the 2019 pandemic (refer to Fig. S3). It was only after the heating season ended that April (73–94) exhibited a downward trend. Comparing the monthly variations in AQI values in Shenyang City from January to April 2020 with those of 2019, we found an inverted U-shaped distribution, with significantly lower AQI values in February and March. In the later stages of the 2021 pandemic (refer to Fig. S4), due to rapid response, precise control measures, and "dynamic clear-out," there was relatively less movement of people and vehicles compared to 2019. As a result, the air quality improved in January (72–100) and February (68–92). However, AQI values in March (94–117) were higher compared to March 2020. As the city gradually resumed work and production in April 2020, pollution sources increased, leading to an upward trend in air quality compared to the same period in 2019 and 2021. The relatively high overall AQI values observed in Shenyang City in January 2020 could be attributed to unfavorable meteorological conditions and increased transportation activities before the Chinese New Year. Hence, the implementation of artificial emission reduction measures during the pandemic control measures resulted in lower AQI levels in the city during February and March 2020.
3.3. Changes in pollutant concentrations before, during, and after the lockdown period
Table 3 provides a summary of the statistical analysis conducted on air pollutants during three different periods. In Fig. 5, we compare the average values of PM2.5, PM10, SO2, NO2, CO, and O3 across these periods. The mean concentration of PM2.5 exhibited a consistent decrease throughout all three periods, with significant differences observed between the pre-lockdown (PRE) and lockdown (LOCK) periods (102.24 ~ 54.03) as well as between the PRE and post-lockdown (POST) periods (102.24 ~ 45.79) (p < 0.05). Similarly, the mean concentration of PM10 initially decreased and then increased over the three periods, with significant differences between the PRE and LOCK periods (134.52 ~ 80.34) and the PRE and POST periods (134.52 ~ 99.10) (p < 0.05). The mean concentration of SO2 followed a pattern similar to PM2.5, with significant differences observed between the PRE and LOCK periods (33.24 ~ 17.66) and the PRE and POST periods (33.24 ~ 15.86) (p < 0.05). Likewise, the mean concentration of NO2 showed a similar trend to PM2.5, with significant differences between the PRE and LOCK periods (60.68 ~ 37.16) and the PRE and POST periods (60.68 ~ 33.34) (p < 0.05). The mean concentration of CO also displayed a similar pattern to PM2.5, with significant differences between the PRE and LOCK periods (1.59 ~ 0.89) and the PRE and POST periods (1.59 ~ 0.74) (p < 0.05). On the other hand, O3 exhibited a monotonic increase across all periods, with significant differences in mean values observed between the PRE and LOCK periods (44.40 ~ 75.59), the LOCK and POST periods (75.59 ~ 111.48), and the PRE and POST periods (44.40 ~ 111.48) (p < 0.05). Notably, there were significant differences in pollutant levels between the lockdown and pre-lockdown periods, and the mean changes in PM2.5, SO2, NO2, and CO were similar.
Table 3
Basic statistics for the air pollutants during PRE, LOOK, and POST.
Pollutant | | Unit | Number | Mean | Std Dev | Min | Max |
PM2.5 | PRE | µg/m3 | 25 | 102.24 | 51.52 | 39 | 265 |
LOCK | µg/m3 | 58 | 54.03 | 41.80 | 11 | 207 |
POST | µg/m3 | 58 | 45.79 | 30.84 | 8 | 131 |
PM10 | PRE | µg/m3 | 25 | 134.52 | 63.30 | 55 | 325 |
LOCK | µg/m3 | 58 | 80.34 | 45.95 | 31 | 237 |
POST | µg/m3 | 58 | 99.10 | 50.57 | 19 | 255 |
NO2 | PRE | µg/m3 | 25 | 60.68 | 19.27 | 29 | 94 |
LOCK | µg/m3 | 58 | 37.16 | 21.20 | 9 | 113 |
POST | µg/m3 | 58 | 15.86 | 8.44 | 4 | 37 |
SO2 | PRE | µg/m3 | 25 | 33.24 | 15.20 | 17 | 74 |
LOCK | µg/m3 | 58 | 17.66 | 7.89 | 7 | 42 |
POST | µg/m3 | 58 | 33.34 | 13.80 | 10 | 72 |
CO | PRE | mg/m3 | 25 | 1.59 | 0.73 | 0.6 | 3.4 |
LOCK | mg/m3 | 58 | 0.89 | 0.48 | 0.2 | 2.8 |
POST | mg/m3 | 58 | 0.74 | 0.29 | 0.3 | 1.5 |
O3-8h | PRE | µg/m3 | 25 | 44.40 | 15.99 | 21 | 75 |
LOCK | µg/m3 | 58 | 75.59 | 14.78 | 52 | 129 |
POST | µg/m3 | 58 | 111.48 | 33.98 | 55 | 191 |
Figure 6 presents the daily average changes in six pollutants PM2.5, PM10, SO2, NO2, CO, and O3) throughout 2020. The daily mean changes in PM and trace gases (SO2, NO2, CO) aligned with the short-term reduction in human activities. In January and early February, the concentrations of particulate matter, SO2, NO2, and CO were significantly higher. This was attributed to the emissions from particulate matter during the winter heating period and the Spring Festival (January 25th and 29th) in Shenyang. Additionally, the low wind speed in January hindered the dispersion of pollutants, leading to high concentrations of particulate matter. In comparison, the lockdown period witnessed a significant reduction in emissions, resulting in a considerable increase of 140.98% and 197.17% in the proportion of days with good air quality (AQI < 100) in Shenyang in March and April, respectively (Table 2). Moreover, the number of days exceeding the second-level standard limits (75µg/m− 3 for PM2.5 and 150 µg/m− 3 for PM10) saw a significant decrease, particularly from February 19th to March 22nd. The daily mean concentrations of SO2, NO2, and CO also experienced significant declines, staying below the second-level standard limits during the pandemic and post-pandemic periods. These three pollutants exhibited relatively stable concentrations from February 19th to March 22nd. The implementation of pandemic control measures led to a noteworthy decrease in particulate matter and emissions of SO2, NO2, and CO from vehicular exhaust. As work resumed gradually on March 23rd, pollutant concentrations began to rise, with the most significant increase occurring in April. Throughout the study period, the daily mean concentration of O3 consistently increased. The study highlights the complexity of ground-level O3 generation, primarily attributed to chemical reactions between nitrogen oxides (NOx) and volatile organic compounds (VOCs) under lighting conditions. The pandemic-induced decline in anthropogenic NOx emissions weakened the "titration" effect (NO + O3→NO2 + O2) resulting from reduced NO concentrations(Liu et al. 2020; Mahato et al. 2020; Sicard et al. 2020; Tang et al. 2022). Additionally, the reduction in fine particulate matter allowed for greater UV penetration, providing more energy for O3 generation at the surface(Tian et al. 2021; Zhou et al. 2016).
3.4 Changes in pollutants during the equivalent lockdown period in 2019 and 2021
Table 4
Basic statistics for air pollutants during LOCK 2019 and 2021.
Pollutant | Unit | Number | Mean | Std Dev | Min | Max |
2019 | | | | | | |
PM2.5 | µg/m3 | 57 | 68.61 | 44.63 | 14 | 201 |
PM10 | µg/m3 | 57 | 110.09 | 55.71 | 30 | 269 |
NO2 | µg/m3 | 56 | 43.50 | 18.59 | 15 | 88 |
SO2 | µg/m3 | 57 | 28.39 | 12.18 | 10 | 57 |
CO | mg/m3 | 57 | 1.10 | 0.55 | 0.4 | 2.5 |
O3-8h | µg/m3 | 57 | 74.47 | 20.40 | 35 | 139 |
2021 | | | | | | |
PM2.5 | µg/m3 | 57 | 64.12 | 45.25 | 14 | 225 |
PM10 | µg/m3 | 57 | 100.35 | 57.64 | 26 | 264 |
NO2 | µg/m3 | 57 | 37.53 | 16.22 | 11 | 81 |
SO2 | µg/m3 | 57 | 18.68 | 7.18 | 8 | 40 |
CO | mg/m3 | 57 | 0.99 | 0.46 | 0.4 | 2.4 |
O3-8h | µg/m3 | 57 | 75.05 | 18.56 | 28 | 127 |
To evaluate the concentration variation patterns of six standard pollutants (2019–2021), the 24-hour pollutant concentrations during the same lockdown period (January 26 to March 23) were used for statistical analysis (Fig. 7). The results are summarized in Table 4. Overall, apart from a slight difference in the mean variation of PM10, the mean variation trends of the other five standard pollutants were similar over the three years. Compared to the unrestricted year of 2019 by the transportation sector, only PM10 and SO2 showed a significant decrease (p < 0.05). Furthermore, compared to the restricted travel year of 2021 with limitations on population and vehicles, there was no significant increase observed in the concentrations of the six standard pollutants. The variation trend in O3 concentration over the three years did not show significant changes and exhibited a monthly increasing pattern, indicating that the variation in O3 concentration was minimally affected by the reduction of emissions during the pandemic period.
Figure 8 Changes in monthly mean PM2.5/PM10, SO2/NO2, PM2.5/SO2and PM2.5/CO ratios for January, February, March, and April 2020 in Shenyang.
To further understand the secondary formation of PM and the variation of pollution sources, the potential main sources of pollutants can be analyzed through the combinations of ratio forms between different pollutant concentrations. Figure 8 shows the monthly average variations of PM2.5/PM10, SO2/NO2, PM2.5/CO, and PM2.5/SO2 from January to April in Shenyang City, with January to March being the central heating period. The monthly average ratio of PM2.5/PM10 exhibited a significant decreasing trend from January (0.77) to April (0.49) (p < 0.05). As shown in Fig. 2, during the lockdown and the later period, there were higher wind speeds, which could lift up dust particles from the surrounding areas and contribute to the higher mass concentration. This indicates a larger contribution of coarse particulate matter pollution, possibly originating from dust storms(Aneja et al. 2001). The monthly average ratio of SO2/NO2 showed significant differences only in January (0.59) and April (0.43) (p < 0.05). The reduction of mobile pollution sources (such as vehicle exhaust emissions) did not result in significant differences in January, February, and March, indicating that the majority of atmospheric pollution came from fixed sources (industrial combustion, etc.)(Song et al. 2017). The monthly average ratio of PM2.5/SO2 showed significant differences in March (2.48) compared to January (3.50) and April (3.84) (p < 0.05), as well as in February (2.77) compared to January (3.50). The lower ratios in February and March indicate that industrial combustion was the main source of PM2.5 during the lockdown period(He et al. 2017). The monthly average ratio of PM2.5/CO showed significant differences in February (51.67) and March (55.59) compared to January (72.96) and April (71.08) (p < 0.05). Similarly, the lower ratios in February and March suggest that the primary pollutants directly emitted from industrial combustion were the main sources of PM2.5 during the lockdown period, rather than secondary pollutants generated through complex physical and chemical reactions involving sulfur dioxide, nitrogen oxides, and volatile organic compounds (VOCs)(Li et al. 2017; Woodruff et al. 2006). The monthly average variations of PM2.5/PM10 and SO2/NO2 in January to April of 2019 (Fig. S5a) and 2021 (Fig. S5b) were similar. Therefore, during the pandemic period, most coal and oil refining enterprises have resumed production to ensure heating for residents. The majority of PM2.5 and trace gases (SO2, NO2, CO) during the lockdown period originated from industrial fixed pollution sources, while the increase in mobile pollution sources after the end of central heating did not result in significant differences in the mean concentrations of PM2.5 and trace gases ((SO2, NO2, CO) between the lockdown and post-lockdown periods (Fig. 5). In addition, compared to the reference periods of 2019 and 2021, the year 2020 saw most people staying at home due to the pandemic lockdown. According to the No. 4 Heating Dispatch Order issued by Shenyang, Liaoning on February 14, in response to the special period during epidemic prevention, large-scale heating networks, central heating power plants, and large-scale thermal power plants such as Shenhai Thermal Power Plant strengthened their operations. As a result, the improvement in air quality during the lockdown period due to the decrease in mobile pollution sources was not as significant (Fig. 7).
3.5 Correlation between air pollutants and meteorological variables
Table 5
Correlation analysis between different atmospheric pollutants and with meteorological factors during PRE, LOCK and POST
| PM2.5 | PM10 | NO2 | SO2 | CO | T(℃) | WS(km/h) | RH(%) |
PRE | | | | | | | | |
PM2.5 | 1 | 0.992** | 0.750** | 0.719** | 0.793** | 0.275 | -0.490* | -0.154 |
PM10 | 0.992** | 1 | 0.748** | 0.733** | 0.816** | 0.255 | -0.479* | -0.167 |
NO2 | 0.750** | 0.748** | 1 | 0.766** | 0.891** | 0.358 | -0.737** | -0.424* |
SO2 | 0.719** | 0.733** | 0.766** | 1 | 0.847** | 0.180 | -0.487* | -0.171 |
CO | 0.793** | 0.816** | 0.891** | 0.847** | 1 | 0.355 | -0.548** | -0.231 |
O3 | -0.604** | -0.600** | -0.872** | -0.772** | -0.792** | -0.430* | 0.716** | 0.451* |
LOCK | | | | | | | | |
PM2.5 | 1 | 0.788** | 0.550** | 0.777** | 0.895** | -0.194 | -0.390** | 0.532** |
PM10 | 0.788** | 1 | 0.380** | 0.659** | 0.714** | -0.100 | -0.115 | 0.268* |
NO2 | 0.550** | 0.380** | 1 | 0.658** | 0.599** | -0.348** | -0.602** | 0.061 |
SO2 | 0.777** | 0.659** | 0.658** | 1 | 0.871** | -0.343** | -0.420** | 0.229 |
CO | 0.895** | 0.714** | 0.599** | 0.871** | 1 | -0.268* | -0.384** | 0.520** |
O3 | 0.082 | 0.119 | -0.214 | 0.091 | 0.037 | 0.461** | 0.255 | -0.175 |
POST | | | | | | | | |
PM2.5 | 1 | 0.785** | 0.813** | 0.705** | 0.817** | -0.225 | -0.007 | -0.398** |
PM10 | 0.785** | 1 | 0.484** | 0.527** | 0.644** | -0.141 | 0.357** | -0.529** |
NO2 | 0.813** | 0.484** | 1 | 0.629** | 0.738** | -0.301* | -0.324* | -0.240 |
SO2 | 0.705** | 0.527** | 0.629** | 1 | 0.897** | 0.031 | 0.107 | -0.232 |
CO | 0.817** | 0.644** | 0.738** | 0.897** | 1 | 0.033 | 0.070 | -0.163 |
O3 | 0.605** | 0.522** | 0.396** | 0.627** | 0.649** | 0.334* | 0.120 | -0.216 |
Note: * and ** mean significant correlation at the P < 0.05 level and highly significant correlation at the P < 0.01 level, respectively |
The study utilized Spearman correlation analysis to assess the relationships between six common air pollutants and three meteorological variables (T, WS, and RH) over three distinct time periods (as presented in Table 5). The results indicated extremely significant positive correlations (p < 0.01) across all three periods for PM and trace gases (SO2, NO2, CO), which suggests similar emission sources such as vehicular exhaust and fossil fuel combustion. Additionally, SO2 and CO demonstrated an extremely significant positive correlation (r > 0.8, p < 0.01) throughout all three study periods. During the pre-lockdown period, industrial combustion and vehicular exhaust were the primary sources of air pollutants. PM2.5 exhibited an extremely significant and strong correlation with PM10 (r = 0.992, p < 0.01). The strong correlations observed between PM and SO2/NO2 (r > 0.7, p < 0.01) were due to these gases being precursors for major components of PM, including SO42− and NO3−. Moreover, the negative correlation between O3 and NO2 (r=-0.872, p < 0.01) was stronger than that of other pollutants because O3 reacts with NO to generate NO2, leading to its depletion. The high concentration of pollutants during the pre-lockdown period also strengthened the negative correlation between O3 and other pollutants, accelerating its consumption. During the lockdown period, industrial combustion was the main source of air pollutants. PM2.5 exhibited a strong positive correlation with CO (r = 0.895, p < 0.01) and a moderate correlation with SO2 (r = 0.777, p < 0.01), indicating that PM2.5 was primarily generated by primary pollutants from industrial combustion. No significant correlations were observed between O3 and other pollutants. After the lockdown period, as centralized heating ended, fixed pollution sources relatively decreased, while mobile pollution sources increased. PM2.5 and trace gases (SO2, NO2, CO) showed extremely significant positive correlations (r > 0.8, p < 0.01). PM2.5 also demonstrated a strong correlation with NO2 (r = 0.813) and CO (r = 0.817). Additionally, significant positive correlations were observed between O3 and other pollutants. The relationship between ozone and other pollutants is often complex, with studies showing that the increase in O3 may cause unintentional changes in secondary PM2.5 due to its enhanced atmospheric oxidation capacity, leading to the formation of SO42−, NO3−, and secondary organic aerosols (SOA).
Wind speed was significantly negatively correlated (p < 0.05) with PM and trace gases (SO2, NO2, CO) during the pre-lockdown period. However, wind speed exhibited an extremely significant positive correlation with O3 (r = 0.716, p < 0.01) during this time. During the lockdown period, wind speed was significantly negatively correlated (p < 0.01) with PM2.5 and trace gases (SO2, NO2, CO) but had no significant correlation with PM10 and O3. After the lockdown period, wind speed showed an extremely significant positive correlation with PM2.5 (r = 0.357, p < 0.01) and a significant negative correlation with NO2 (r=-0.324, p < 0.05). Temperature only showed a significant negative correlation (r=-0.430, p < 0.05) with O3 during the pre-lockdown period. During the lockdown period, temperature was significantly negatively correlated (p < 0.05) with trace gases (SO2, NO2, CO) and extremely significantly positively correlated with O3 (r = 0.461, p < 0.01). After the lockdown period, temperature was significantly negatively correlated with NO2 (r=-0.301, p < 0.05) and significantly positively correlated with O3 (r = 0.334, p < 0.05). Relative humidity exhibited significant correlations with NO2 (r=-0.424, p < 0.05) and O3 (r = 0.451, p < 0.05) during the pre-lockdown period. During the lockdown period, it showed significant positive correlations with PM2.5 (r = 0.532, p < 0.01), PM10 (r = 0.268, p < 0.05), and CO (r = 0.520, p < 0.01). After the lockdown period, it was significantly negatively correlated with PM2.5 (r=-0.398, p < 0.01) and PM10 (r=-0.529, p < 0.01).
3.6 Backward trajectory analysis
To track the sources and transmission pathways of three air masses (PRE, LOCK, POST) in Shenyang during 2020, we plotted backward trajectories (Fig. 8). During the PRE period, we identified four clusters originating from different wind directions. Clusters 1 (29%), 2 (35%), and 3 (24%) primarily came from the northwest and dominated the transport direction. Cluster 1 represented long-range transport originating from Hulunbuir, Inner Mongolia, passing through the Songliao Basin, the western part of the Horqin Sandy Land, the Greater Khingan Range, the Horqin Sandy Land, and the Northeast China Plain. Clusters 2 and 3 originated from the northeast and east of Inner Mongolia, respectively. Cluster 4 (12%) represented short-range regional transport from the southwest, originating from the Yellow Sea. During the LOCK period, clusters 1 (32%), 2 (18%), 3 (13%), and 4 (19%) also came from the northwest and north, dominating the transport direction (82%). There was no significant difference compared to the PRE period in terms of transport direction. In the case of long-range regional transport, the proportion of Cluster 4 from eastern Mongolia decreased, and a pollution trajectory from the southeast of Russia (Cluster 3) appeared. Cluster 1 originated in the southwest of Heilongjiang and was transmitted to Shenyang along the west side of the Northeast China Plain. Cluster 4 originated from the northwest of Liaoning Province, while Cluster 5 (19%) arrived in Shenyang from the northeast of Shandong Province through the Yellow Sea. After the lockdown ended, there was a significant change in the trajectory's transition direction. Compared with the LOCK period, the proportion of Clusters 4 (13%) and 3 (14%) from Russia, Mongolia, and Shandong significantly decreased. Cluster 5 (19%) represented a long-range pollution trajectory from the border between northern Heilongjiang and Inner Mongolia. Cluster 1 (30%) originated in northeastern Inner Mongolia, and Cluster 2 (25%) came from the Yellow Sea and was transmitted to Shenyang through Dalian, Anshan, and Liaoyang.
To determine if there were unique transport pathways for pollutants entering Wuhan, we conducted similar trajectory analyses for three periods in 2019 without lockdown measures and compared the results with those of 2020. During the PRE and Reference periods in 2019, the transport direction was mainly northwest and southwest (Fig. S1), showing no significant difference compared to the PRE and LOCK periods in 2020. In the PRE period of 2019, Clusters 4 (17%) and 5 (21%) from Mongolia and Russia, respectively, as well as Clusters 1 (26%) and 2 (19%) from Inner Mongolia, originated from the northwest. Cluster 3 (17%) from Huludao, Liaoning Province, came from the southwest. In the Reference period of 2019, Clusters 1 (24%) and 2 (23%) from Inner Mongolia, and Cluster 5 (22%) from Mongolia, came from the northwest. Cluster 3 (16%) from Huludao, Liaoning Province, and Cluster 4 (15%) from Shandong Province, came from the southwest. During the POST period in 2019, Cluster 2 (12%) and 3 (31%) came from the intersection of eastern and northern Inner Mongolia and Russia, while Cluster 1 (13%) and 5 (31%) respectively came from Tianjin and Qingdao, Shandong Province.
Comparing the results of the LOCK period in 2020 with the Reference period in 2019, there was no significant difference in the transport direction of pollutants entering Shenyang. Russia implemented comprehensive epidemic prevention policies on March 30, 2020, including production shutdown and home isolation. Therefore, pollution trajectories from the direction of Russia were observed in Fig. 8(a). Mongolia reported its first case of COVID-19 on March 2, 2020, and announced traffic blockade in Ulaanbaatar and provincial capital cities until March 16, 2020. The proportion of pollution trajectories from Mongolia during the LOCK period decreased relatively. Inner Mongolia issued a level one epidemic prevention and control alert at the beginning of January, which was adjusted to level three at the end of February. The rapid recovery of production and resulting air pollution caused Clusters 1 and 2 to shift towards Inner Mongolia during the LOCK period. Additionally, the average wind speed during the lockdown was 8.14 km/h, with a maximum wind speed of 17 km/h occurring more frequently in northerly winds (Fig. 2), greatly promoting the cross-regional spread of pollutants.